TLR9−/− mice show low bone mass and increased osteoclastogenesis

To investigate the role of TLR9 in osteoclastogenesis and bone metabolism, we first assessed the bone mass of TLR9−/− (KO) and wild-type (WT, TLR9+/+) mice with a C57BL6/J genetic background36,37 by micro-CT analysis. Interestingly, we found that TLR9−/− mice presented significantly lower trabecular bone mass in both the distal femur and vertebrae (Figs. 1a–c and S1a, b). There was no difference in cortical bone between the two groups (Fig. S1c–f). TLR9−/− mice also showed elevated levels of bone resorption (CTX) and formation (P1NP) markers in their circulation, which indicates a high bone turnover status in TLR9−/− mice (Fig. 1d). Furthermore, histomorphometric analysis showed significantly increased osteoclast numbers and higher bone resorption indices in TLR9−/− mice (Fig. 1c, e). Although there was no significant difference in osteoblast numbers or surface area between the two groups, the mineral apposition rate (MAR) was higher in TLR9−/− mice (Fig. S1g–j). Real-time PCR using calvarial osteoblasts from TLR9−/− and WT mice showed that the expression of Alp and Osx was higher in the TLR9−/− group, whereas there were no significant differences in other osteoblastic genes (Fig. S1k). The alkaline phosphatase (ALP) assay after in vitro osteoblast differentiation showed a slightly higher level of ALP in TLR9−/− osteoblasts, but the level of statistical significance was not reached (Fig. S1l, m).

Fig. 1

TLR9−/− mice had low bone mass, increased osteoclastogenesis and systemic inflammation. a Representative 3D μCT images of femurs and L3 vertebrae from TLR9−/− (KO) and wildtype (WT) mice. b Trabecular bone microarchitecture of femurs and L3 vertebrae showing bone mineral density (BMD), bone volume per tissue volume (BV/TV), number of trabeculae (Tb. N), trabecular thickness (Tb.Th) and trabecular separation (Tb.Sp). KO, n = 8; WT, n = 9. c Representative HE (upper panels) and TRAP (lower panels) stained images of distal femurs from each group. d Circulating CTX and P1NP levels. KO, n = 6; WT, n = 8. e Histomorphometric analysis of bone resorption indices. n = 8 per group. f–h In vitro osteoclastogenesis assay. f TRAP-stained osteoclast like cells (OCLs). g Quantification of number and area of TRAP+ multinucleated OCLs. h qPCR analysis of osteoclast signature genes’ expression at the end of assay. i Western blots showing expression of the indicated molecules in early osteoclast precursors. j–k In vitro osteoclastogenesis of wildtype BMNCs in the presence of TLR9 antagonist. j TRAP-stained OCLs. k Quantification of OCL numbers. l Spleens from 8-week-old male TLR9−/− and wildtype mice. m Levels of TNFα, IFNγ, IL1β, and RANKL in serum and bone marrow supernatants. n = 3 to 13 per group. n Splenic T cell populations analyzed by flow cytometry. n = 6–10 per group. o The bone marrow T cell populations. n = 5–10 per group. The numbers in n and o represent the frequencies in total splenic or bone marrow cells. Eight-week-old sex-matched mice were used in d, e and m–o. One-way ANOVA with Turkey’s multiple comparison tests was conducted in k. In other panels, statistical significance was determined using an unpaired two-tailed t-test. Error bars represent the s.d. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.000 1 and ns P > 0.05

Next, we assessed the role of TLR9 in osteoclast differentiation by conducting an in vitro osteoclastogenesis assay. Primary nonadherent bone marrow mononuclear cells (BMNCs) from TLR9−/− and WT mice were stimulated with RANKL and CSF1 for 5 days. We found that the number of osteoclast-like cells (OCLs) was substantially increased in the TLR9−/− group, with no significant change in the area per OCL (Fig. 1f, g). Quantitative PCR results showed that TLR9−/− OCLs (Day 5) presented higher expression of osteoclast signature genes, including CTSK, RANK and NFATc1, than WT OCLs (Fig. 1h). Western blot analysis also revealed that the protein levels of RANK, Fos, NFATc1 and CSF1R were upregulated in TLR9−/− OCPs on Day 2 of in vitro culture (Fig. 1i). Furthermore, the TLR9−/− OCPs showed a higher proliferation rate than the WT cells in response to CSF1 (Fig. S1n). These results indicate that TLR9−/− bone marrow cells possess a greater potential to differentiate into osteoclasts than WT cells.

To verify whether the loss of osteoclast-intrinsic TLR9 was the main reason for the increase in osteoclastogenesis in TLR9−/− mice, we cultured WT BMNCs in the presence of a TLR9-specific antagonist (ODN-2088, 0-10 μmol·L−1) to investigate whether the molecular blockade of TLR9 promotes osteoclastogenesis to a similar extent to the genetic ablation of TLR9. Our results showed that the TLR9 antagonist had no significant effect on osteoclast differentiation; however, we observed a suppressive effect on osteoclastogenesis when the TLR9 antagonist was applied at a very high concentration (10 μmol·L−1) (Fig. 1j, k). This result indicates that increased osteoclastogenesis in TLR9−/− mice may not be caused by the loss of osteoclast-intrinsic TLR9 signaling. Alternatively, the increase in osteoclastogenesis in TLR9−/− mice may be mediated by other mechanisms.

TLR9−/− mice exhibit chronic systemic inflammation

To further explore the mechanism underlying osteoclastic bone loss in TLR9−/− mice, we first measured the levels of pro-osteoclastic cytokines in the circulation. Strikingly, the levels of inflammatory cytokines with potent osteoclastogenic activity, including TNFα, IL1β and RANKL, but not IL17 and IL6, were significantly elevated in serum samples from TLR9−/− mice (Figs. 1m and S2a, b). In contrast, circulating OPG levels were decreased, which led to a significant increase in the RANKL/OPG ratio in TLR9−/− mice (Fig. S2b). The circulating level of IFNγ, an important activator of both innate and adaptive immune responses, was also increased in TLR9−/− mice (Fig. 1m). In line with the results obtained for circulating cytokines, higher TNFα, IL1β and RANKL levels and a lower OPG level were also found in the bone marrow supernatants of TLR9−/− mice (Figs. 1m and S2b). In addition to higher levels of inflammatory cytokines in the circulation, TLR9−/− mice exhibited enlarged spleens relative to their WT counterparts (Fig. 1l). These results suggest the presence of systemic inflammation in TLR9−/− mice.

To investigate whether there were any changes in immune cells and which immune cells contributed to the increase in inflammatory cytokines seen in TLR9−/− mice, we phenotyped immune cells from the bone marrow and spleen of TLR9−/− and WT mice by flow cytometry. Our results showed that the TLR9−/− mice presented increased proportions of splenic CD4+ and CD8+ T cells (Figs. 1n, o and S2c, d). We also observed significantly increases in bone marrow CD4+ T cells, with a trend of an increase in bone marrow CD8+ T cells in TLR9−/− mice. Notably, TLR9−/− mice showed more CD69+-activated CD4+ and CD8+ T cells than the WT mice in both their bone marrow and spleen (Figs. 1n, o and S2c, d). Intracellular cytokine analysis revealed that the numbers of TNFα- and IFNγ-producing CD4+ T cells were markedly increased in the bone marrow and spleen of TLR9−/− mice (Fig. 1n, o). The TLR9−/− mice presented more TNFα- and IFNγ-producing CD8+ T cells than the WT mice in the spleen but not in the bone marrow (Fig. S2c, d). There were also increased numbers of RANKL-producing splenic CD8+ T cells and TNFα-producing macrophages in the TLR9−/− mice (Fig. S2c). In accordance with the flow cytometry results, TNFα, IFNγ and RANKL levels and the RANKL/OPG ratio were elevated in the culture supernatant of TLR9−/− splenic CD4+ T cells (Fig. S2e). Thus, our findings indicate that CD4+ T cells are important contributors to the increased levels of inflammatory cytokines seen in TLR9−/− mice.

B lymphocytes show a close relationship with bone cells, as the development and maturation of B cells take place in bone marrow. Therefore, we also investigated whether TLR9 deficiency affects B cells in bone marrow and in peripheral lymphoid tissue. Interestingly, we found that bone marrow and splenic B cell counts were decreased in the absence of TLR9 (Fig. S2c, d). B cells are RANKL and OPG producers, and mature B cells are the major source of OPG in bone marrow.4 To determine the levels of B cell-derived RANKL and OPG, we cultured purified splenic and bone marrow CD19+ B cells from TLR9−/− and WT mice and measured secreted RANKL and OPG in the culture supernatant. Similar to CD4+ T cells, TLR9−/− B cells showed increased production of soluble RANKL but a decreased level of OPG in the supernatant (Fig. S2f). As osteoblasts and bone marrow mesenchymal cells (BMSCs) are also important sources of RANKL and OPG, we further cultured osteoblasts and BMSCs from TLR9−/− and WT mice. No differences in the levels of osteoblast- and BMSC-derived RANKL and OPG were found between the culture supernatants harvested from TLR9−/− and WT cells (Fig. S2g). These results suggest that B cells also contribute to the imbalance in the RANKL/OPG ratio after the deletion of TLR9.

Systemic inflammation is the major cause of increased osteoclastogenesis and bone loss in TLR9−/− mice

To investigate whether the inflammatory cytokines produced by immune cells were responsible for the increase in osteoclastogenesis in TLR9−/− mice, we cultured splenocytes from TLR9−/− and WT mice together with BMNCs from Rag1−/− mice using a Transwell system. Rag1−/− mice are devoid of T and B lymphocytes but show no change in bone mass relative to WT mice;38 thus, the use of an ideal source of BMNCs without T and B cell contamination and a Transwell culture system allows only soluble factors to flow freely. We found that the soluble factors produced by TLR9−/− splenocytes induced a significantly higher count of osteoclasts from BMNCs than WT splenocytes (Fig. 2a, b). Furthermore, coculturing TLR9−/− splenic CD4+ T cells, but not TLR9−/− splenic B cells, with Rag1−/− BMNCs significantly promoted osteoclast differentiation (Fig. S3a–d). This result further suggests that CD4+ T cells are the major contributors to increased osteoclastogenesis in TLR9−/− mice. To confirm these in vitro results in vivo, we performed adoptive cell transfer experiments, in which splenocytes from TLR9−/− or WT mice were adoptively transferred into Rag1−/− recipient mice. Intriguingly, Rag1−/− mice that received TLR9−/− splenocytes (KO_Rag1−/− mice) showed a lower bone mass than those that received WT splenocytes (WT_Rag1−/− mice) (Figs. 2c, d and S3e). Accordingly, CTX level was also higher in KO_Rag1−/− mice, but without any change in the P1NP level (Fig. 2e), indicating that the bone loss of KO_Rag1−/− mice was due to increased bone resorption by osteoclasts.

Fig. 2

Systemic inflammation in TLR9−/− mice plays an important role in osteoclastic bone loss. a, b Rag1−/− BMNCs were cocultured with TLR9−/− (KO) or wildtype (WT) splenocytes and stimulated with RANKL and CSF1 in a transwell system. The OCLs were stained by TRAP (a) and numbers quantified (b). c–j Adoptive transfer of TLR9−/− splenocytes induced bone loss in Rag1−/− mice. c Representative 3D μCT images of femurs from each group. d Trabecular BMD, BV/TV, and Tb. N of femurs and L3 vertebrae in recipient mice (KO_Rag1−/−, n = 10; WT_Rag1−/−, n = 11). e Serum CTX and P1NP. KO_Rag1−/−, n = 7; WT_Rag1−/−, n = 8. f Spleens of KO_Rag1−/− and WT_Rag1−/− mice. g–h Spleen (g) and bone marrow (h) T cell populations analyzed by flow cytometry. n = 4–6 mice per group. The numbers represent the frequencies in total splenic or bone marrow cells. i Serum cytokine levels. KO_Rag1−/−, n = 7; WT_Rag1−/−, n = 8. j–p The bone marrow transplantation experiment. k Trabecular BMD, BV/TV, and Tb. N of femurs and L3 vertebrae in the recipient mice (KOchim, n = 6; WTchim, n = 7). l–m In vitro osteoclastogenesis assay using BMNCs from KOchim and WTchim mice. l TRAP-stained OCLs. m Quantification of OCL numbers. n Circulating CTX and P1NP levels. KOchim, n = 7; WTchim, n = 8. o Serum TNFα, IL6 and RANKL levels. KOchim, n = 7; WTchim, n = 8. p Spleens of KOchim and WTchim mice. Error bars represent the s.d. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.000 1 and ns P > 0.05; statistical significance was determined using an unpaired two-tailed t-test

Furthermore, the adoptive transfer of TLR9−/− splenocytes recapitulated the inflammatory status in TLR9−/− mice. KO_Rag1−/− mice exhibited larger spleens than WT_Rag1−/− mice (Fig. 2f) and showed increased counts of total CD4+ and CD8+ T cells as well as CD69+-activated CD4+ and CD8+ T cells in the bone marrow and spleen (Figs. 2g, h and S3f, g). Similar to TLR9−/− mice, KO_Rag1−/− mice exhibited elevated levels of TNFα, IFNγ, IL1β, RANKL and IL6 in their circulation (Figs. 2i and S3h). TNFα- and RANKL-producing CD4+ T-cell counts were markedly increased in the spleens of KO_Rag1−/− mice (Figs. 2g and S3g). There were also more IFNγ-producing CD4+ and CD8+ T cells in the bone marrow of KO_Rag1−/− mice (Figs. 2h and S3f). Consistent with the findings in TLR9−/− mice, we found increased counts of total macrophages and TNFα-producing macrophages in the spleens of KO_Rag1−/− mice (Fig. S3g). These observations further indicate that the splenocytes of TLR9−/− mice are able to cause inflammation and bone loss in recipient mice. Together, these data show that the inflammation induced by immune cells plays an important role in mediating osteoclastic bone resorption in TLR9−/− mice.

In addition to inflammation in the circulation and spleen, our data suggest that inflammation also occurred in the bone marrow of TLR9−/− mice, since upregulated inflammatory cytokine levels in the bone marrow supernatant and an increased number of activated bone marrow T cells were observed. To verify the contribution of bone marrow inflammation and the hematopoietic compartment to bone loss in TLR9−/− mice, we constructed bone marrow chimeras by transferring TLR9−/− and WT bone marrow cells to 6-week-old lethally irradiated WT mice. Bone marrow chimerism was confirmed by the PCR genotyping of sex chromosomes in bone marrow cells (Fig. S3i). Six weeks after cell transfer, we found that mice that received TLR9−/− bone marrow cells (KOchim) presented a significantly lower bone mass in the femur and spine than the mice that received WT bone marrow cells (WTchim) (Figs. 2j, k and S3j). KOchim bone marrow cells showed an increased osteoclastogenic potential in vitro (Fig. 2l, m). Accordingly, KOchim mice presented a significantly elevated level of CTX, whereas the level of P1NP was similar between the two groups (Fig. 2n). These results indicate that osteoclastic bone resorption is the major cause of bone loss in KOchim mice. Furthermore, similar to TLR9−/− mice, KOchim mice also exhibited enlarged spleens, increased circulating TNFα, IL6 and RANKL levels and decreased circulating OPG levels (Figs. 2o, p and S1k). A trend toward higher circulating IL1β and IFNγ levels was also observed in TLR9−/− mice (Fig. S3k). These results demonstrate that the transplantation of TLR9−/− bone marrow cells promoted bone loss and a low level of systemic inflammation in recipient mice, further indicating that bone marrow inflammation and hematopoietic cells play an important role in osteoclastic bone loss in TLR9−/− mice.

The secretion of proinflammatory cytokines is one of the basic elements of inflammation,39 and one key feature of the systemic inflammation observed in TLR9−/− mice is increased levels of multiple pro-osteoclastogenic inflammatory cytokines. Among these cytokines, TNFα is a central player in inflammatory bone loss,1 and TLR9−/− mice showed significantly increased TNFα levels in the circulation and bone marrow. To determine the role of TNFα in bone loss in TLR9−/− mice, we treated WT and TLR9−/− mice with an anti-TNFα antibody or control IgG. In line with previous reports showing that the blockade of TNFα decreased bone resorption,40 we found that the trabecular bone volume of TLR9−/− mice was significantly increased after anti-TNFα treatment and became similar to the bone volume of IgG-treated WT mice (TNFα_KO versus IgG_WT) (Figs. 3a, b and S4a). Anti-TNFα therapy also resulted in a significantly decreased CTX level, with no significant change in the P1NP level in TLR9−/− mice (Fig. 3c). Interestingly, circulating IL6 and TNFα levels in TLR9−/− mice were reduced after anti-TNFα therapy (Fig. S4b). We also observed trends toward lower circulating RANKL and IFNγ levels in TLR9−/− mice after anti-TNFα therapy (Fig. S4b). Furthermore, anti-TNFα therapy decreased the frequency of splenic CD4+ and CD8+ T cells in TLR9−/− mice (Fig. S4c). The restoration of bone mass by anti-TNFα therapy suggests an important role of TNFα in the bone loss of TLR9−/− mice. Taken together, the above results indicate that the chronic systemic inflammation observed in TLR9−/− mice is the major cause of increased osteoclastogenesis and subsequent bone loss.

Fig. 3

The anti-TNFα therapy experiment and the altered gut microbiota composition and intestinal inflammation in TLR9−/− mice. a–c, l, m The anti-TNFα therapy experiment. In all panels, n = 3 in IgG_KO and anti-TNFα_WT group; n = 4 in anti-TNFα_KO and IgG_WT group. a Representative 3D μCT reconstructions of femurs from each group. b Trabecular BMD, BV/TV, and Tb. N of femurs in each group. c Serum CTX and P1NP levels. d 3D PCoA of 16S rRNA sequencing of fecal microbiota from 8-week sex-matched TLR9−/− (KO) and wildtype (WT) mice. Each dot represents a fecal microbiota from one mouse. KO, n = 12; WT, n = 14. e–f Relative abundance of the significantly upregulated families (P < 0.05) in KO (e) and WT (f) group. Boxplots represent median and quantiles. g Cladogram of Linear discriminant analysis Effect Size (LEfSe) analysis showing the differentially abundant bacteria between KO and WT groups. h Circulating LPS and IgA levels. n = 8–10 per group. i Large intestines of 8-week-old male KO and WT mice. j Fecal Lcn-2 levels. KO, n = 5; WT, n = 6. k Flow cytometry analysis of T cell populations in MLN and PP. n = 6 per group. The numbers represent the frequencies in total MLN or PP cells. l, m Ameliorated gut inflammation after anti-TNFα therapy. l Large intestines from all treated mice. m Serum LPS and fecal Lcn-2 levels. Mann–Whitney t test was used in e and f. Two-way ANOVA with multiple comparisons (Turkey’s test) were used in b, c and m. An unpaired two-tailed t-test was applied in other panels. Error bars represent the s.d. *P < 0.05, **P < 0.01, ****P < 0.000 1 and ns P > 0.05

TLR9 deficiency alters the gut microbiota composition and induces intestinal inflammation

Although our data demonstrated chronic inflammation resulting in bone loss in TLR9−/− mice, the source of this inflammation was not clear. TLR signaling has been shown to affect gut microbial ecology, and an altered gut microbial composition is an important driving force in inflammation and diseases.41,42 Recent studies have further suggested that the gut microbiota plays a key role in inflammatory bone loss.7,43 We therefore envision that TLR9 deficiency may alter the gut microbial composition, which could be the primary source of inflammation.

To test our hypothesis, we first performed 16S rRNA sequencing of the microbiome using feces from WT and TLR9−/− mice. Principal coordinate analysis (PCoA) of the microbiomes of these mice showed segregation between the two groups (Fig. 3d). Both the richness of the gut flora and Shannon’s diversity were similar between WT and TLR9−/− mice (Fig. S4e). Relative abundance at the family level showed differences in the composition of the microbiota between the two groups (Fig. S4f). We observed a TLR9−/− microbiota signature associated with an increase in the families Deferribacteraceae, Odoribacteraceae, Rikenellaceae and Staphylococcaceae and a decrease in the families Erysipelotrichaceae and Turicibacteraceae (Fig. 3e, f). Notably, Deferribacteraceae, Odoribacteraceae and Rikenellaceae, which are gram-negative bacteria, presented higher abundance in TLR9−/− mice than in WT mice. This observation was further supported by the linear discriminant analysis (LDA) effect size (LEfSe) results (Figs. 3g and S4g). Interestingly, the phylum Deferribacteres was specifically enriched in the TLR9−/− group (Figs. 3g). An analysis of the ten most abundant species showed that two gram-negative bacteria, Mucispirillum schaedleri (M. schaedleri, of the phylum Deferribacteres) and Parabacteroides distasonis (P. distasonis, of the phylum Bacteroidetes), presented significant increases in TLR9−/− mice (Fig. S4h). Interestingly, both M. schaedleri and P. distasonis have been reported to be associated with IBDs.44,45 Taken together, our results showed that the gut microbiota was altered such that pathologic properties were increased in the absence of TLR9.

Next, we investigated whether the altered gut microbiota in TLR9−/− mice affected systemic immunity. To this end, we tested the circulating levels of IgA and lipopolysaccharide (LPS), an endotoxin produced mainly by gram-negative bacteria. Interestingly, both circulating LPS and IgA levels were higher in TLR9−/− mice (Fig. 3h). The increased LPS level was in accordance with the higher abundance of gram-negative bacteria in TLR9−/− mice. We also assessed the length of the large intestine from the ileocaecal junction to the anus in the mice. Intriguingly, the large intestine was significantly shorter in TLR9−/− mice than in their WT counterparts (Fig. 3i), which indicates intestinal inflammation46 in TLR9−/− mice. We then measured fecal Lipocalin 2 (Lcn-2), a marker of colitis, and found that TLR9−/− mice presented higher levels of fecal Lcn-2 than WT mice (Fig. 3j). Next, we analyzed immune cells in mucosal-associated lymphoid tissues, and a higher proportion of CD4+ T cells was observed in the mesenteric lymph nodes (MLNs) and Peyer’s patches (PPs) of TLR9−/− mice (Fig. 3k). TLR9−/− mice also showed increased CD8+ T cell counts in their PPs (Fig. 3k). Furthermore, after treatment with anti-TNFα, the length of the large intestine of the TLR9−/− mice was increased to the same length observed in the WT mice (Fig. 3l). Anti-TNFα treatment also significantly reduced the levels of fecal Lcn-2 and circulating LPS (Fig. 3m) in TLR9−/− mice. CD4+ and CD8+ T cell counts were significantly decreased in the PP of TLR9−/− mice after treatment with anti-TNFα (Fig. S4d). Together, these results suggest the presence of subclinical chronic intestinal inflammation in TLR9−/− mice.

An altered gut microbiota plays an important role in systemic inflammation and subsequent bone loss in TLR9−/− mice

To verify whether the altered gut microbiota contributes to the inflammation and changes in bone mass observed in the absence of TLR9, germ-free (GF) TLR9−/− and WT mice were rederived via hysterectomy and maintained in gnotobiotic facilities for 8 weeks. Consistent with our hypothesis, the trabecular bone mass of GF TLR9−/− mice was restored to the same level observed in their WT counterparts at the age of 8 weeks (Figs. 4a, b and S5a). Although the P1NP level was still higher in GF TLR9−/− mice, the CTX level of GF TLR9−/− mice was similar to that of GF WT mice (Fig. 4c) and was markedly lower than that of conventional TLR9−/− mice (Fig. 1d). An in vitro osteoclastogenesis assay using BMNCs from GF TLR9−/− and WT mice also showed a similar number of OCLs between the two groups (Fig. 4d, e).

Fig. 4

Altered gut microbiota played an important role in the systemic inflammation and bone loss in TLR9−/− mice. a–i Phenotypic study of 8-week-old male germ-free (GF) TLR9−/− and wildtype mice. a Representative 3D μCT images of the femurs. b Trabecular BMD, BV/TV, and Tb. N of femurs and L3 vertebrae in each group. GF_KO, n = 4; GF_WT, n = 6. c Circulating CTX and P1NP levels. GF_KO, n = 4; GF_WT, n = 6. d–e In vitro osteoclastogenesis using BMNCs from the GF mice. d TRAP-stained OCLs. e Quantification of OCL numbers. f Circulating cytokine levels. GF_KO, n = 4; GF_WT, n = 6. g Spleens from each group. h–i Flow cytometry analysis of splenic CD4+ T cells and macrophages (h) and bone marrow CD4+ T cells (i). n = 5 per group. j–r The cohousing experiment. j Representative 3D μCT images of the femurs. k Trabecular BMD, BV/TV, and Tb. N of femurs and L3 vertebrae in male Co-H_KO and Co-H_WT mice. n = 7 per group. l Serum CTX and P1NP levels. n = 11 per group. m TNFα and IFNγ levels. Serum, n = 14 per group; BM supernatant, n = 3 per group. n Large intestines from male Co-H_KO and Co-H_WT mice. o Serum LPS and fecal Lcn-2 levels. LPS, n = 3 per group; Lcn-2, n = 8 per group. p, q Spleen (p) and bone marrow (q) CD4+ T cell populations by flow cytometry. n = 6 per group. Twelve-week-old sex-matched mice were used in l, m and o–q. The numbers in h, i, p and q represent the frequencies in total splenic or bone marrow cells. Significance was determined using an unpaired two-tailed t-test. Error bars represent the s.d. *P < 0.05, **P < 0.01 and ns P > 0.05

Then, we compared the inflammatory status of GF TLR9−/− and WT mice. With the exception of a higher level of circulating RANKL in GF TLR9−/− mice, no significant differences in the levels of other cytokines were observed between the two groups (Fig. 4f). Notably, the circulating levels of TNFα, IL1β and IFNγ were markedly decreased in GF TLR9−/− mice compared to TLR9−/− mice raised under conventional conditions (Figs. 4f and 1m). Furthermore, the size of the spleen was reduced in GF TLR9−/− mice and was similar to that observed in GF or conventional WT mice (Figs. 1l and 4g). No significant difference in the proportions of splenic CD4+ and CD8+ T cells, macrophages or CD69+-activated T cells was found between the two GF groups (Figs. 4h and S5b). Although there was a slight increase in bone marrow CD8+ T cells in GF TLR9−/− mice, the frequencies of other bone marrow T cell populations were similar between the two groups (Figs. 4i and S5b). Although GF TLR9−/− mice presented fewer splenic and bone marrow B cells (Fig. S5b), the difference in the proportion of bone marrow B cells between GF TLR9−/− and WT mice was significantly decreased compared to the difference between conventional TLR9−/− and WT mice (Fig. S2d). Together, these results showed decreased bone resorption along with lower inflammation levels in GF TLR9−/− mice, further suggesting that the gut microbiota is an important source of inflammation in TLR9−/− mice.

Cohousing can facilitate microbial transfer among individual animals.47 To gain further support for findings obtained in GF mice, we cohoused newly weaned WT and TLR9−/− mice for 9 weeks. By sequencing the gut microbiota from fecal samples of cohoused mice, we found that the differences in the gut microbiota between the two genotypes were normalized after cohousing (Fig. S5c–h). The abundances of the signature TLR9−/− families were similar between the two cohoused groups, with an increase in Deferribacteraceae in the WT mice and a decrease in Odoribacteraceae and Rikenellaceae in the TLR9−/− mice after cohousing (Fig. S5f). At the species level, the abundances of both M. schaedleri and P. distasonis were similar between the two cohoused groups (Fig. S5h). Thus, our results confirmed that cohousing TLR9−/− and WT mice facilitated microbial transfer and normalized the difference in the gut microbiota between the two groups.

Next, by investigating the effect of cohousing on the bone phenotype, we discovered that the trabecular bone density of TLR9−/− and WT mice became similar after cohousing (Figs. 4j, k and S5i–k). The circulating CTX level in cohoused TLR9−/− mice was decreased to a level similar to that in the cohoused WT mice, and no difference in P1NP levels was found between these two groups (Fig. 4l). Cohousing with WT mice also lowered TNFα and IFNγ levels in the circulation and bone marrow of TLR9−/− mice, and no difference in the levels of circulating and bone marrow TNFα, IL6, IFNγ and OPG was found between the cohoused TLR9−/− and WT mice (Figs. 4m and S5l). Although RANKL and IL1β levels were still higher in the cohoused TLR9−/− mice than in their WT counterparts (Fig. S5l), their levels in the bone marrow of cohoused TLR9−/− mice were significantly lower than those in the bone marrow of single-housed TLR9−/− mice (Fig. 1m). Interestingly, cohousing also normalized spleen size and large intestine length between the TLR9−/− and WT mice (Figs. 4n and S5m). Fecal Lcn-2 and serum LPS levels were similar between the two cohoused groups (Fig. 4o). No significant differences in the T cell populations of the MLNs, PPs, spleen and bone marrow were observed between TLR9−/− mice and WT mice after cohousing (Figs. 4p–q and S5n–p). Taken together, the above results are in line with the findings in GF mice, further suggesting that the alteration of the gut microbiota is critical for the development of inflammation and subsequent bone loss in TLR9−/− mice.

Myeloid-biased hematopoiesis plays an important role in inflammatory bone loss in TLR9−/− mice

Although our data showed that immune cell-produced inflammatory cytokines promoted osteoclastogenesis in TLR9−/− mice in vivo, they did not fully explain why TLR9−/− BMNCs generated more osteoclasts than WT cells in vitro after induction with the same concentrations of osteoclastogenic cytokines. Recently, emerging evidence has shown that chronic inflammation disturbs the normal homeostasis of hematopoiesis and results in myeloid skewing.48,49 Since TLR9−/− mice showed increased inflammatory cytokines and fewer B lymphocytes in their bone marrow, we hypothesize that the chronic inflammation observed in TLR9−/− mice may result in myeloid-biased hematopoiesis, which may further promote osteoclastogenesis by increasing the frequency of OCPs in the bone marrow.

Bone marrow is a highly heterogeneous tissue comprised of many cell types. To further dissect the effect of TLR9 on the development of different cell lineages, we employed single-cell RNA sequencing (scRNA-seq) to examine differentially expressed genes (DEGs) in BMNCs from TLR9−/− and WT mice using the 10x Genomics Chromium platform. Following rigorous quality control, we compiled gene expression data for clustering analyses from 36 263 cells (18 664 and 17 599 cells from TLR9−/− and WT mice, respectively). This revealed 32 distinct populations visualized as uniform manifold approximation and projection (UMAP) embeddings. Population nomenclature was based on specific gene expression, and we identified 10 clusters of B cells, 5 clusters of macrophages/monocytes and precursors, 5 clusters of neutrophils, 2 clusters of erythrocytes and precursors, and 1 cluster of each of the following cell types: T cells, dendritic cells, NK cells, monocyte-dendritic cell precursors (MDPs), HSPCs, mesenchymal cells, basophils, plasma cells and megakaryocytes (Figs. 5a and